Interest segmentation of hyperspectral imagery

Abstract

In recent years, many new methods for analyzing spectral imagery have been introduced. These new methods have been developed to improve the analysis of hyperspectral imagery. Many of these techniques are data driven anomaly/target detection and spectral clustering algorithms which are used to decide whether a particular pixel or area is “interesting.” For this research, a group of these algorithms are used on two tiled hyperspectral images. The results of each algorithm are combined into a multi-band feature image. The features are combined in such a way that the image is segmented into regions that either contain “interest” or do not.